Re: [commit+7.6] [patch+7.6] Fix 7.5 regression crashing GDB if gdbserver dies

[Pedro Alves <palves at redhat dot com>]

On 03/22/2013 08:42 PM, Jan Kratochvil wrote:
> On Fri, 22 Mar 2013 20:53:56 +0100, Pedro Alves wrote:
>> Debugging a live production system with kgdb?  I don't buy
>> that, really.  If one cares to debug the kernel with kgdb, then
>> one can just as easily patch the kernel, just like one patches the
>> kernel for security bugs and the like.
> 
> This is becoming offtopic but the production systems I face are isolated from
> Internet and no security patches get ever applied.  One of the many reasons is
> to keep all the systems the same, with interchangeable hard drives.
> 

I'll eat my socks when you prove to me you've needed to debug one of
those with a kgdb that is actually affected by this bug.  ;-)

> 
>>> People still face for example the bug
>>> 	new gdbserver is incompat. with old gdb - ;core: suffix
>>> 	http://sourceware.org/bugzilla/show_bug.cgi?id=15230
>>> which is "fixed" in GDB (also) since 2010:
>>> 	commit 3d972c80ab566c07f5e55d356821fb883c9ade88
>>> 	Date:   Tue Jan 12 21:40:23 2010 +0000
>>>
>>
>> I've said before that I regret not having noticed that problem
>> at patch review time or before the release.  I still do.  I don't
>> see the parallel here though. We can't go back in time and fix older
>> GDBs to cope with ;core...  Users can upgrade their GDBs though.
> 
> In practice in such cases people rather stay with older application, for
> whatever reason.

Then if those people will prefer staying with older gdb and older
gdbserver, fixing either gdb or gdbserver would do absolutely
nothing for them.  We can't rewrite history for them.

>> Because it's a workaround that does not need to exist.  It adds
>> maintenance burden in the wrong place.  The time we've wasted
>> collectively iterating on this shows it.
> 
> Otherwise people would need to carry the workarounds downstream.  So far
> I think upstream GDB is supporting workarounds of broken system components.

Yes, when necessary.  This was one case where my belief is it was not.

But yes, let's move on.  That old broken kgdb will get older over
time, and at some point not even you will care.  :-)

>> I like this version much better than the original.  If the exception
>> swallowing ever turns out problematic again, I'll propose reverting
>> the original patch again.  So in interest of moving forward, this one's
>> fine with me.
> 
> Checked in:
> 	http://sourceware.org/ml/gdb-cvs/2013-03/msg00199.html
> and for 7.6:
> 	http://sourceware.org/ml/gdb-cvs/2013-03/msg00200.html

Thanks.  I do believe the new exception will come in handy, so
I'm glad to see that go in.

-- 
Pedro Alves